AI SEO, explained without the hype
AI search does not rank ten links. It writes one answer and names a handful of sources. This is how you become one of them.
AI SEO is the practice of making a website easy for AI search systems to read, trust and quote. It is not a separate channel from SEO. The same crawlable pages, clear structure and genuine expertise still decide who wins; the difference is that AI systems extract passages rather than list links, so pages that answer a question directly, in specific and well-structured language, get cited while pages that circle the topic get skipped.
What AI SEO actually means
Three things changed at once. Google started answering questions above the results with AI Overviews. ChatGPT, Perplexity, Claude and Copilot became places people ask before they search. And both of those surfaces cite a small set of sources rather than listing a page of blue links.
That is the whole shift. You are no longer competing for a position in a list. You are competing to be the passage a model chooses to summarise. The industry has invented several names for working on this, and they broadly describe the same job:
- AI SEO, the umbrella term for optimising towards AI-mediated search.
- AEO, answer engine optimisation, focused on formatting content so a direct answer can be lifted from it.
- GEO, generative engine optimisation, focused on being cited inside generated answers. Note that GEO still means geographic targeting in older material.
- LLM visibility or AI search optimisation, usually the same work sold under a newer label.
If a supplier presents these as four separate products with four separate fees, that is a packaging decision, not a technical one.
How AI search picks its sources
Nobody outside the model providers can see the exact selection logic, and it changes. But the observable behaviour is consistent enough to build against.
Most assistants either retrieve live results from a search index or work from a crawl of the web, then generate an answer grounded in a few retrieved passages. That means two gates. First, can the system fetch and parse your page at all. Second, does one passage on that page answer the question so cleanly that quoting it is the path of least resistance.
Passing the first gate is technical: server-rendered HTML, fast responses, no critical content locked behind JavaScript, no robots rules blocking AI crawlers you actually want. Passing the second is editorial: say the answer in the first two sentences, then support it.
What makes a page citable
Citable pages share a shape. They front-load the conclusion, they use headings that read like the questions people ask, and they put the specifics in a structure a machine can parse rather than burying them in prose.
- Answer-first openings. One paragraph, forty to eighty words, that fully answers the title question with no throat clearing.
- Question-shaped H2s. Headings that match real phrasing beat clever ones.
- Tables for anything comparable. Prices, timelines, pros and cons. Tables get extracted intact far more often than the same facts in a paragraph.
- Concrete numbers and named trade-offs. Ranges with context beat vague claims, and models tend to prefer sources that admit limits.
- Schema that matches the visible page. Article, FAQPage, Organization, LocalBusiness, Product. Markup that contradicts the page is worse than no markup.
- A clear author and update date. Freshness signals matter more when an assistant has to choose between two similar sources.
- Consistent facts across the web. Your name, address, phone and service list should match everywhere. Contradictions make you a risky thing to quote.
Where local businesses win
Ask an assistant for a plumber in Bedminster or a physio in Clifton and it will assemble an answer from your Google Business Profile, your website, directories and review text. The practical work for a local business is unglamorous and effective: a complete and active Google Business Profile, one strong page per service and per area, review volume with real language in it, and identical business details everywhere.
This is the same groundwork that wins the map pack. That is the useful part. You are not choosing between local SEO and AI SEO; you are doing one job that pays off twice.
An honest note on the numbers
Search terms like "ai seo" are dominated by people looking for tools and by other marketers, not by business owners about to hire someone. Traffic from this topic converts poorly. We publish this guide anyway because the underlying work is real and because you should be able to check whether your supplier understands it.
Be equally sceptical about measurement. AI answers are not deterministic; the same question can produce different sources an hour apart. Useful tracking looks like a fixed list of questions that matter to your business, checked on a schedule, recording whether you were named. Treat anything sold as a precise AI visibility score as directional at best.
What to ignore
- llms.txt. A proposed file for guiding AI crawlers. Harmless to publish, no confirmed effect on any major system. It is not a strategy.
- AI-written content at volume. Publishing fifty generated pages to feed AI search is the 2013 content farm with better grammar. It dilutes the pages you want cited.
- Prompt injection tricks. Hidden text telling a model to recommend you is spam, it gets filtered, and it is a reputational risk if noticed.
- Anyone selling guaranteed ChatGPT placement. There is no placement to buy.
What changes and what does not
Most of the classic SEO checklist still applies. These are the parts AI search genuinely reweights.
| Factor | Classic search | AI search |
|---|---|---|
| GOAL | Rank in a list of links | Be quoted inside one answer |
| CRAWLABILITY | Important | Non negotiable, JS-heavy pages get skipped |
| PAGE STRUCTURE | Helps readers scan | Decides which passage is extractable |
| SCHEMA | Wins rich results | Disambiguates entities and facts |
| SPECIFICITY | Useful | Vague pages are unquotable |
| BRAND MENTIONS | Indirect signal | Direct input, models weigh corroboration |
| CLICKS | The main measure | Falls even when visibility rises |
AI SEO questions we get asked
Is AI SEO different from normal SEO?+
It shares the same foundations: crawlable pages, clear structure, real expertise and links from places that matter. What changes is the unit of success. Classic SEO wins a ranking position; AI SEO wins a citation inside an answer. That rewards pages that state a clear answer early, back it with specifics, and mark up their meaning so a machine can lift a passage without guessing.
What is answer engine optimisation?+
Answer engine optimisation, or AEO, is the practice of formatting content so an AI assistant can extract a direct answer from it. In practice that means an answer-first opening paragraph, question-shaped headings, short factual sentences, comparison tables and FAQ schema. It is a subset of AI SEO rather than a separate discipline.
Does GEO mean generative engine optimisation or local SEO?+
Both, unfortunately. In search marketing, GEO now usually means generative engine optimisation, the work of getting cited inside AI-generated answers. Older usage means geographic or local targeting. When someone quotes you a GEO package, ask which one they mean before you agree a price.
Can you guarantee my business appears in ChatGPT or AI Overviews?+
No, and anyone who does is guessing. AI answers are generated fresh each time and vary by phrasing, location and model version. What you can control is whether your site is technically eligible to be cited: crawlable, structured, specific and current. We build for that eligibility and we track mentions, but we do not sell guarantees on systems nobody can see inside.
Do AI search visitors actually convert?+
Fewer of them arrive, and they arrive later in the decision. AI answers absorb the research clicks, so raw traffic often falls while enquiry quality rises. Judge AI SEO on enquiries and on how often you are named in answers, not on sessions in Google Analytics.
Is a WordPress site a problem for AI search?+
Not inherently, but the common failure modes cluster there: plugin-generated markup that contradicts the page, heavy JavaScript that delays the real content, and thin pages spun up by a content plugin. A clean, fast, well-marked-up build removes those problems by default. Our guide on WordPress versus coded sites covers where WordPress is still the right call.